Re: hostap TX fix in 5.x [Fwd: Re: wi hostap speed]

From: Scott Pilz <scottp_at_tznet.com>
Date: Sat, 15 May 2004 09:42:42 -0500 (CDT)
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1


	Brian, try this exact setup on 5.x - this is for the hostap side,
the client side can be anything compatible. . .

	The card is of course, a PRISM 2.5 chipset. To be more specific, a
Demarc PRISM 200mw PCMCIA (www.demarc.com). In case there is any question,
I've also tried non Demarc cards (still PRISM 2.5) ranging from 30mw to
200mw. The hardware varies - from a P133 to a AMD1.8, doesn't make any bit
of difference.

	I've tried nearly all the public firmware releases for this
chipset. The one I'm on right now (that I'll use in this example) Primary
1.1.1, Station 1.7.4.. See output below:

dmesg output:

<snip>
wi0: 802.11 address: 00:60:b3:..:..:..
wi0: using RF:PRISM2.5 MAC:ISL3873
wi0: Intersil Firmware: Primary (1.1.1), Station (1.7.4)
wi0: 11b rates: 1Mbps 2Mbps 5.5Mbps 11Mbps
</snip>

	The setup of the system doesn't seem to matter either. I have
tried with BRIDGE(4) enabled between the interfaces, I've tried setting
the system up to be a router (and route the packets to someplace else),
and just for testing, I've tried not having the system route packets or
bridge packets, instead just ftp directly from the associated client to
the hostap station itself - the speeds do not change. Everything works -
the bridging, the routing, the ftping to the station, but the speeds are
less than half of what they should be during downloads - upload FROM
client TO station works at full 500+KB/sec.

(in this example, I'll use 192.168.0.1 for the hostap station and .2 for
the client)

<snip>
ifconfig 192.168.0.1 netmask 255.255.255.0 media DS/11Mbps mediaopt hostap
ssid test channel 1 stationname test-system
</snip>

	Obviously the channel will be up to you, as well as the IP address
of the system. I've also used media DS/2Mbps as well as media autoselect -
DS/11Mbps gets the better speeds typically (well, what you could call
'better' compared to the other settings, but not full speeds like in
FBSD-4.x).

	I tested this with WEP, long long ago, but I don't currently run
it. The type of traffic does not matter. I have tried FTP, I've tried java
web-based "speed test" type software, I've tried downloading in browsers
from associated stations, again - it doesn't matter. The speeds do not
change. At average, the maximum speed is about 160-170KB/sec.

	Again, using this exact same setup line-for-line, booting with BSD
4.x (and throwing in older firmware as 4.x does not support this PRISM 2.5
firmware) gets full 500+KB/sec up and down.

	I've gone as far as .. I want to say, flash version 1.3.6 on the
PRISM 2.5 card, no change at all.

	If you figure this out, you're better than I am :)...Short of
re-writing the wi driver I'm out of ideas.

	Scott

On Sat, 15 May 2004, Brian Fundakowski Feldman wrote:

> Scott Pilz <scottp_at_tznet.com> wrote:
> > 	Brian, speeds are limited to under 180KBytes/sec while in hostap
> > mode. Take any 5.x box with the wi drivers and a prism 2.5 based card (in
> > order for hostap to work properly). Set it up to be an access point
> > (mediaopt hostap, doesn't matter what you use for media, I have used both
> > autoselect, 11Mbps, etc.) Use any radio equipment to connect to the access
> > point (from Lucent to Tranzeo to SmartBridges, you name it, doesn't matter
> > as long as it's 802.11b compatible obviously).. You'll get a maximum
> > DOWNLOAD of about 180Kbps if you're lucky.. Upload works fine at
> > 500+Kbps.. Now, take that exact same setup without changing anything on a
> > 4.9 system - download speeds (to the client radio) are back up to 500+
> > Kbps.
> >
> > 	This is just off the top of my head, but there was some major
> > reconstruction in wi(4) from 4.9 to 5.x. It's been broken since day one
> > since the first release of 5.x.. so looking at 'what changed' in CVS
> > probably won't help you. My guess is that it's not going to be easily
> > found. Bless anyone who wants to look into this - as this has been a real
> > nightmare.
>
> What are you testing with to get those figures?  I will try this out at work
> on Monday where I have umpteen boxes and PRISM 2.x cards if you can give me
> even more exact reproduction steps.  If you don't say anything I'm just
> going to try netperf TCP stream tests on a non-WEP hostap network.
>
> --
> Brian Fundakowski Feldman                           \'[ FreeBSD ]''''''''''\
>   <> green_at_FreeBSD.org                               \  The Power to Serve! \
>  Opinions expressed are my own.                       \,,,,,,,,,,,,,,,,,,,,,,\
>
>
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.4 (FreeBSD)

iD8DBQFApixj2REUg6gjWxgRAtW7AJ4/4+VQJKk4aNZ47CLP7lts/s0RegCg85di
m1EXEpWGTMQFuSklbXngzYo=
=z2vm
-----END PGP SIGNATURE-----
Received on Sat May 15 2004 - 05:42:58 UTC

This archive was generated by hypermail 2.4.0 : Wed May 19 2021 - 11:37:54 UTC